修改探针逻辑

app/services/ai_tool.py

@@ -72,38 +72,44 @@ async def ai_chat(llm: BaseChatModel, system_prompt: str, user_message: str, sce
             if not _gate_probing:
                 # 当前协程成为探针，负责退避重试直到成功
                 _gate_probing = True
-                result = await _probe_until_recover(llm, system_prompt, user_message)
-                # 记录 AI 调用日志
-                if scene:
-                    await _save_call_log(scene, system_prompt, user_message, result)
-                # 探针成功，直接返回结果（不用再调一次）
-                return result
+                await _probe_until_recover(llm, system_prompt, user_message)
+                # 探针恢复后，回到 while 循环顶部用业务数据重新调用
+                continue
             # 已有探针在工作，回到 while 顶部 await _gate_event.wait() 等待唤醒
 
 
 async def _probe_until_recover(llm: BaseChatModel, system_prompt: str, user_message: str) -> str:
-    """探针：固定间隔重试直到接口恢复，返回成功的响应内容，并打开门闸唤醒所有等待协程"""
+    """探针：用轻量 hello 消息探测接口是否恢复，成功后打开门闸，原协程重新走正常流程"""
     global _gate_probing
 
-    delay = 0.5  # 固定间隔 500ms
+    delay = 0.2  # 探测间隔 200ms
     attempt = 0
 
-    while True:
-        await asyncio.sleep(delay)
-        attempt += 1
-        try:
-            messages = [
-                SystemMessage(content=system_prompt),
-                HumanMessage(content=user_message),
-            ]
-            response = await llm.ainvoke(messages)
-            # 成功 → 打开门闸，唤醒所有等待协程
-            _gate_probing = False
-            _gate_event.set()
-            log.info("AI 接口恢复，门闸打开，第{}次探测成功", attempt)
-            return response.content
-        except Exception as e:
-            log.warning("AI 接口仍不可用，第{}次探测失败(间隔{}ms): {}", attempt, int(delay * 1000), e)
+    try:
+        while True:
+            await asyncio.sleep(delay)
+            attempt += 1
+            try:
+                log.info("探针第{}次探测开始...", attempt)
+                messages = [
+                    SystemMessage(content="You are a helpful assistant."),
+                    HumanMessage(content="hello"),
+                ]
+                await llm.ainvoke(messages)
+                # 成功 → 打开门闸，唤醒所有等待协程
+                _gate_probing = False
+                _gate_event.set()
+                log.info("探针第{}次探测成功，AI 接口恢复，门闸打开", attempt)
+                break
+            except Exception as e:
+                log.warning("探针第{}次探测失败: {}", attempt, e)
+    except (asyncio.CancelledError, Exception) as e:
+        # 探针意外死亡 → 必须恢复门闸状态，否则所有 worker 永远阻塞
+        _gate_probing = False
+        _gate_event.set()
+        log.error("探针协程意外退出，强制打开门闸: {}", e)
+
+    return ""
 
 
 async def ai_chat_json(llm: BaseChatModel, system_prompt: str, user_message: str, scene: Optional[str] = None) -> Any: